Basel Committee on Banking Supervision. Consultative Document. Revisions to the minimum capital requirements for market risk

Basel Committee on Banking Supervision Consultative Document Revisions to the minimum capital requirements for market risk Issued for comment by 20 June 2018 March 2018

This publication is available on the BIS website (www.bis.org). Bank for International Settlements 2018. All rights reserved. Brief excerpts may be reproduced or translated provided the source is stated. ISBN 978-92-9259-150-2 (online)

Contents Revisions to the minimum capital requirements for market risk... 1 Introduction... 1 1. Standardised approach... 1 1.1 Revisions to the treatment of liquid FX pairs... 2 1.2 Revisions to correlation scenarios... 2 1.3 Revisions to capital requirements for non-linear instruments... 3 1.4 Revisions to risk weights... 4 1.5 Other clarifications... 4 2. Internal models approach... 5 2.1 P&L attribution test... 5 2.1.1 PLA test input data... 5 2.1.2 PLA test metric design... 6 2.1.3 PLA test failure consequences... 7 2.1.4 Trading desk requirements... 8 2.2 Non-modellable risk factors... 8 2.2.1 Process for satisfying modellability requirements and expectations for internal model calibration... 9 2.2.2 Impact of the NMRF framework on seasonal markets... 10 2.2.3 Impact of NMRF idiosyncratic equity risk... 10 3. Scope of market risk capital requirements... 11 3.1 Treatment of structural FX positions... 12 3.2 Boundary between the trading book and the banking book... 12 4. Simplified alternative to the standardised approach... 13 Next steps... 13 Annex A Revisions to the standardised approach... 15 A.1 Revisions to correlation scenarios... 15 A.2 Revisions to the curvature risk capital requirement... 15 A.3 Revisions to FX risk factors and curvature risk capital requirement... 17 A.4 Treatment of multi-underlying options and index instruments... 17 A.5 Revisions to the treatment of liquid FX pairs... 18 A.6 Revisions to standardised approach risk weights for GIRR, equity and FX risk classes... 19 Annex B Revisions to the internal models approach... 20 Revisions to the minimum capital requirements for market risk iii

B.1 Revisions to PLA test metric design... 20 B.2 Revisions to risk factor modellability... 21 B.3 Revisions to the IMA capital requirement and PLA test failure consequences... 24 B.4 Revisions to Appendix B: PLA test metric design... 25 B.5 Revisions to Appendix B and Glossary: PLA definitions... 29 Annex C Revisions to trading desk structure... 30 Annex D Guidance for evaluating the sufficiency and accuracy of risk factors for IMA trading desk models... 31 Annex E Revisions to the scope of market risk capital requirements... 35 E.1 Revisions to the treatment of structural FX positions... 35 E.2 Revisions to the boundary between the trading book and banking book... 35 Annex F Simplified alternative to the standardised approach to market risk capital requirements... 39 iv Revisions to the minimum capital requirements for market risk

Revisions to the minimum capital requirements for market risk Introduction In January 2016, the Basel Committee on Banking Supervision published the standard Minimum capital requirements for market risk 1 (hereafter January 2016 standard ). This new market risk standard was developed to address a number of structural shortcomings in the Basel II market risk framework (and its subsequent revisions), and served as a key component of the Basel Committee's reform of global regulatory standards in response to the global financial crisis. In the time since its publication, the Basel Committee has monitored the pace of implementation of the market risk standard as well as its impact on banks market risk capital requirements. In acknowledgment of ongoing challenges related to implementation of the standard, the Basel Committee s oversight body, the Group of Governors and Heads of Supervision (GHOS), has endorsed an extension of the implementation date to 1 January 2022 (which will constitute both the implementation and regulatory reporting date for the standard). This deferred implementation date is intended to allow banks additional time to develop the systems infrastructure needed to apply the standard and for the Committee to address certain specific outstanding issues. In order to address the issues with the standard that the Committee has identified, this consultative document proposes a number of revisions to the standard. It also sets out the Committee s proposals for a simplified alternative to the revised standardised approach to market risk, which take into account responses to the consultative document the Committee issued in June 2017. 2 1. Standardised approach A major structural shortcoming of the Basel II market risk framework is that it does not feature a risksensitive standardised approach that can serve as a credible fallback for, as well as a floor to, the internal models approach. The January 2016 standard intended to address this by introducing a revised standardised approach. The main element of the revised standardised approach the Sensitivities-based Method relies on the use of sensitivities. Sensitivities are banks estimates of how much the values of their financial instruments change when the values of a prescribed list of underlying risk factors change. For example, banks are required to calculate the change in value of their financial instruments if there was a 1 basis point move in interest rates. The standardised approach specifies: the risk weights that should be applied to the sensitivities for each of the prescribed list of risk factors. Banks mutliply their sensitivities to risk factors by these risk weights to estimate the change, on a risk factor by risk factor basis, in the value of their trading book portfolio; and the approach that banks should use to aggregate the risk factor-level valuation changes into an aggregate amount that is the basis of the capital requirement a set of formulae is prescribed that uses defined correlation assumptions to provide diversification benefit across risk factors. 1 Basel Committee on Banking Supervision, Minimum capital requirement for market risk, January 2016, www.bis.org/bcbs/publ/d352.pdf. 2 Basel Committee on Banking Supervision, Consultative Document Simplified alternative to the standardised approach to market risk capital requirements, June 2017, www.bis.org/bcbs/publ/d408.pdf. Revisions to the minimum capital requirements for market risk 1

The use of sensitivities and the incorporation of diversification benefits in calculating the aggregate capital requirement better align the outcomes of the revised standardised approach with that of the internal models approach by enabling a degree of risk sensitivity in the standardised approach. The Committee s ongoing monitoring of the impact of the revised standardised approach, and feedback received from banks as they have begun to implement it, have highlighted areas where the approach to measure risk factor-level losses, and their aggregation, are not commensurate with the actual risk. Without revision, these issues could make the standardised approach a less credible fallback for the internal models approach. The Committee therefore proposes revisions to the following elements: the approach to determine FX pairs that are liquid and therefore subject to lower risk weights; 3 the correlation scenarios applied in the standardised approach calculation; and the treatment of non-linear financial instruments such as options. In order to ensure that the overall level of capital requirements resulting from the revised standardised approach is more consistent with the Committee s initial expectation, the Committee is also proposing reductions in the risk weights applied for certain asset classes. Further details on the proposed revisions are provided below, with revisions to the associated standard text provided in Annex A. 1.1 Revisions to the treatment of liquid FX pairs Under both the standardised approach and the internal models approach, certain specified currency pairs are designated to be sufficiently liquid to warrant lower associated capital requirements. Neither approach recognises that it is possible to combine two liquid currency pairs to create a new, triangulated pair that, by virtue of being the result of combining two liquid instruments, would also be liquid. For example, although USD/BRL and USD/EUR are included in the January 2016 standard s list of liquid currency pairs, EUR/BRL is not included in the list. However, by combining two liquid instruments that reference USD/BRL and USD/EUR, a bank could create a liquid instrument that references the currency pair EUR/BRL. Because the January 2016 standard does not permit the consideration of such combinations, some liquid FX currency pairs may be subject to capital requirements that are not commensurate with their risk. The Committee proposes to allow banks to combine two currency pairs in the current list of liquid pairs and treat the resulting new FX pair as liquid. 1.2 Revisions to correlation scenarios In the January 2016 standard, improved risk sensitivity in the revised standardised approach relative to the Basel II standardised approach is largely due to better recognition of diversification benefits across banks trading portfolios. These diversification benefits come from the use of prescribed correlations in the aggregation of risk factor-level losses to calculate banks capital requirements. To account for a range of possible market conditions, the January 2016 standard requires banks to calculate capital requirements for each risk class 4 three times: (i) with correlation assumptions as prescribed in the standard (the medium correlations scenario); (ii) with all correlations scaled upwards by 25% (the high correlations scenario); and (iii) with all correlations scaled downwards by 25% (the low 3 The proposed revision to the treatment of FX liquidity would also apply to the internal models approach. 4 The standardised approach categorises market risks into seven risk classes and calculates capital requirements separately for each. The defined classes are: (i) general interest rate risk; (ii) equity risk; (iii) credit spread risk: non-securitisation; (iv) credit spread risk: securitisations (non-correlation trading portfolio); (v) credit spread risk: securitisations (correlation trading portfolio); (vi) commodity risk; and (vii) foreign exchange risk. 2 Revisions to the minimum capital requirements for market risk

correlations scenario). The capital requirement for each risk class is calculated in each scenario, with the ultimate capital requirement determined by the highest result, aggregated across the entire trading book portfolio, from the three scenarios. For risk factors that are observed empirically to be consistently highly correlated in all market conditions, the Committee has observed that the low correlations scenario can produce correlations that are more conservative than empirical data would support. This can make the outcome of the standardised approach overly conservative. The Committee proposes to revise the low correlations scenario to address this issue by limiting the reduction in correlations in these cases. 1.3 Revisions to capital requirements for non-linear instruments The January 2016 standard specifies additional capital requirements curvature risk capital requirements for certain financial instruments, such as options, for which values do not change linearly with respect to their underlying risk factors. Because of this non-linearity, these types of instruments can lose more value than would be estimated by applying shocks to risk factor sensitivities. The curvature risk capital requirements are computed by calculating the maximum loss of two scenarios of shocks an upward shock and a downward shock. Banks revalue their non-linear instruments based on those shocks, and calculate the incremental value change beyond what would be estimated using sensitivities. This incremental amount is the additional capital requirement for curvature risk. The Committee has identified three aspects of the curvature risk measurement where minor changes could improve the January 2016 standard: The approach to apply shock scenarios: when calculating the curvature risk capital requirement, the upward and downward shocks are applied separately to each risk factor. The worst loss for each risk factor is used to calculate the capital requirement. This approach can lead to two financial instruments that are very closely related having capital requirements based on different shocks. The Committee proposes to revise this approach so that consistent scenarios are applied to risk factors that are defined to be in the same standardised approach bucket 5 for the credit spread risk, equity and commodity risk classes. The Committee is also exploring an alternative approach of defining sectors as a subset of each bucket and applying consistent scenarios at that level and would welcome feedback on the potential merits and drawbacks of this alternative approach. Cliff effects caused by the approach used to calculate aggregate capital requirements: the Committee has observed that the formulae used to calculate the aggregate curvature risk capital requirement can cause cliff effects for certain types of trading book portfolios. Cliff effects arise from the use of an alternative specification that banks must use when curvature risk positions are negative (ie when banks would only see profits in the curvature shock scenarios). The alternative specification can lead to an abrupt increase in capital requirements. To address this, the Committee proposes a simple fix that applies a floor to the part of the formula causing the cliff effect. Potential double-counting of FX curvature risk: the revised standardised approach requires banks to define FX exposures relative to their reporting currency. This reflects the reality that a bank s FX risk stems from movements in other currencies relative to that which it uses for reporting purposes. However, the Committee has observed that, in the specific situation of banks holding FX options where neither of the underlying currencies is the bank s reporting currency, the 5 Within the January 2016 standard standardised approach, buckets are defined groups of risk factors with similar characteristics. Revisions to the minimum capital requirements for market risk 3

approach to calculate curvature risk capital requirements may lead to double-counting. 6 The Committee has not received sufficient data through its monitoring to determine whether this double-counting of FX curvature risk is a material issue in practice. The Committee therefore seeks feedback via this consultation on whether this is a material issue. If this is a material issue, a potential revision to the standard to address it is set out in Box 1. The Committee welcomes views on whether the approach set out below would address the issue in an appropriate manner. In providing feedback on this matter, commenters should provide concrete evidence and data in support of any recommendations for the appropriate level of any scaling factor. Box 1 131. For FX and equity curvature risk factors, the curvature risk weights are relative shifts ( shocks ) equal to the delta risk weights. For FX curvature, where none of the underlying currencies of a particular FX instrument is the reporting currency, any resulting curvature sensitivities may be divided by a scalar [X]. If a bank opts to apply this discretion, it must do so consistently for all FX instruments where none of the underlying currencies is the reporting currency. 1.4 Revisions to risk weights The Committee s monitoring of the impact of the January 2016 standard indicates that the currently reported capital impact of the revised standardised approach is not consistent with its initial expectations. The Committee has identified that, after taking into account the expected impact of the above-proposed revisions, reductions in risk weights in the January 2016 standard standardised approach are necessary to bring market risk capital requirements closer to that originally intended level. Based on impact data received to date, the Committee proposes to reduce the risk weights for the general interest rate risk class by 20 40%, and equity and FX risk classes by 25 50%. No specific revision is proposed to the risk weights applied for the credit spread and commodity risk classes. The final recalibration for all risk classes will be determined based on further analysis of impact data provided by banks, and feedback provided to this consultative document. Upon finalisation of any recalibrated risk weights, the Committee may also consider making corresponding changes to risk weights used in the standardised approach to credit valuation adjustment risk (SA-CVA) given that SA-CVA risk weights were based upon the risk weights included in the January 2016 market risk standard. 1.5 Other clarifications The Committee has received feedback that the treatment of multi-underlying options and index instruments in the revised standardised approach is unclear. The Committee proposes, in Annex A, revisions to clarify the treatment. 6 For example, if a GBP-reporting bank writes an option on the EUR/USD exchange rate, the bank is considered to have two separate FX risk exposures a GBP/EUR exposure and a GBP/USD exposure. The bank calculates curvature risk capital requirements based on two shocks: one where EUR is shocked relative to GBP, and one where USD is shocked relative to GBP. This is unlike the case of an EUR-reporting bank, which will have one FX risk exposure and calculates curvature risk capital requirements based on one shock: where USD is shocked relative to EUR. 4 Revisions to the minimum capital requirements for market risk

2. Internal models approach The internal models approach (IMA) set out in the January 2016 standard featured a number of enhancements relative to the Basel II framework, including: (i) enhanced requirements for approval to use models, including the introduction of a profit and loss (P&L) attribution (PLA) test that trading desks must pass on an ongoing basis in order to be eligible for the IMA; and (ii) more coherent and comprehensive methods to measure risk, including the introduction of distinct capital requirements for non-modellable risk factors (NMRFs). To address a number of issues identified in the course of ongoing monitoring of the implementation of these aspects of the IMA and to facilitate its effective implementation, the Committee proposes the following revisions. 2.1 P&L attribution test The Committee introduced the PLA test to serve as an objective, quantitative assessment of whether models that a bank uses to calculate market risk capital requirements appropriately measure all material risks relevant to each individual trading desk to which they are applied. The test is intended to be a benchmark by which supervisors can assess the appropriateness of a bank s use of a model, with a trading desk being required to use the standardised approach for market risk in the event that it fails to meet the PLA test s requirements. The PLA test specified in the January 2016 standard compares historical time series of two measures of daily P&L for each trading desk for which a bank intends to use internal models: Hypothetical P&L (HPL): The P&L, as calculated by the bank s systems that produce the reported daily P&L but removing commissions, fees, the impact of intraday trading and certain valuation adjustments. Risk-theoretical P&L (RTPL): The P&L that is produced when only the risk factors in the bank s internal risk management model, and the valuation techniques used in that model, are included. The HPL is the benchmark against which the PLA test assesses a trading desk s risk management model. If the risk management model includes all risk factors and uses the same valuation techniques as used by the models used for the reported daily P&L, then the RTPL will match the HPL. The two P&Ls might differ, however, in the event that the bank s risk management model ignores some risks, and/or values products in a simpler way, in order to allow the model to run more efficiently. Therefore, material inconsistencies between the two measures of P&L are indicative of missing risks that are not included in the risk management model. The PLA test aims to measure and set a limit on how inconsistent the two P&Ls can be before a trading desk is no longer permitted to use the IMA. The Committee continues to believe that the PLA test is an important innovation of the market risk framework, but is aware of issues with the approach specified in the January 2016 standard that may lead to it not meeting the Committee s intended objectives. These issues relate to the inputs to the test, the test metrics themselves, and the automatic consequences of failing the test. To address these issues, the Committee proposes a number of revisions that are described below. Associated revisions to the standard text are in Annex B. 2.1.1 PLA test input data The Committee has received requests for additional clarity on the January 2016 standard s definitions of HPL and RTPL, in addition to requests for clarification as to whether banks may be permitted to make any adjustments to data used to produce both parameters. Revisions to the minimum capital requirements for market risk 5

As explained above, the objective of the PLA test is to assess the materiality of risks that may be missing from the risk management model due to risk factors that are not included in the model or simplifications in the model s approaches to valuation. Beyond these sources of discrepancy between the HPL and the RTPL of a given trading desk, additional differences between the two measures of P&L may arise as the result of acknowledged differences or misalignments in the data that the bank uses as inputs to calculate each measure. For example, such differences could be the result of: differences in the times at which market data are collected to calculate the HPL and the RTPL, respectively; and/or a bank using different data providers to source the inputs for its calculations of the HPL and the RTPL. To avoid issues resultant from such differences in input data, the Committee proposes revisions to the definitions of the HPL and the RTPL and clarification that banks may align input data. The proposed revisions specify conditions that a bank must meet in making such alignments to ensure that the process does not inappropriately conceal the impact of any missing risks that the PLA test is intended to assess. 2.1.2 PLA test metric design The PLA test as specified in the January 2016 standard includes two test metrics to measure the size of the difference between the HPL and the RTPL for each trading desk, 7 with each test metric to be calculated on a monthly basis using data over the previous one-month period. Banks must then count the number of breaches of the test over the previous 12 one-month periods. The Committee has monitored the performance of the PLA test metrics as originally specified, and has identified concerns over the metrics combined abilities to appropriately identify models that demonstrate deficiencies for the purposes of determining capital requirements. In addition, the onemonth sample of data used may be insufficiently representative of a model s performance. To address these concerns, the Committee proposes revisions to: (i) the frequency at which the test is to be conducted and the length of the time series to be used; and (ii) the design of the test metrics themselves. The Committee proposes that the PLA test be calculated on a quarterly basis using a time series of data collected over the preceding 12 months. The Committee also proposes two new metrics to replace those specified in the January 2016 standard. The first of the two new proposed test metrics would measure the correlation between the values of the HPL and the RTPL, comparing the Spearman correlation with predefined thresholds. The second of the proposed test metrics would measure the similarity of the distributions of the HPL and the RTPL. For use as this second test metric, the Committee seeks feedback on two alternatives. The first alternative would use the Kolmogorov-Smirnov (KS) test, whereas the second alternative would use a Chi-squared test. The details of these tests are summarised in Box 2. The Committee seeks comments on the relative merits of the KS and the Chi-squared test alternatives, particularly with regard to considerations over their ease of implementation and ability to discriminate between appropriate and inappropriate models for a given trading desk. 7 The first test metric measures the mean of unexplained daily P&L (ie daily RTPL minus daily HPL) over the standard deviation of HPL, and the second metric measures the variance of unexplained P&L over the variance of HPL. 6 Revisions to the minimum capital requirements for market risk

Box 2 Overview of the proposed PLA test metrics The proposals to revise the PLA test metrics are intended to provide reliable measures of whether the P&L generated by a risk management model (represented by the RTPL) is sufficiently related to and similar to the P&L generated by the bank s front office (represented by the HPL). To measure the relationship between the two measures, the Committee proposes that the revised test assess whether the correlation between the HPL and the RTPL is sufficiently high using the Spearman correlation metric. To measure how similar the measures are, the Committee proposes two alternatives for the revised test to assess whether the time series of values of the HPL and the RTPL have sufficiently similar statistical distributions. Assessing the relationship between the HPL and the RTPL: Spearman correlation The Spearman correlation coefficient is a measure of the correlation between the two time series of P&Ls to assess the level of dependence between the HPL and the RTPL. The metric separately ranks (from lowest value to highest) the historical 12-month time series of daily HPL and RTPL values. The correlation between the assigned ranks of the two series is calculated. A strong correlation will only be observed if the rank ordering of values is closely related between the two time series. A well modelled trading desk would be expected to exhibit a strong correlation. Assessing the similarity of the distributions of the HPL and the RTPL The measurement of similarity of the distributions of the measures of P&L is intended to assess whether the risk management model used for the trading desk sufficiently captures the P&Ls of the desk across the range of market conditions during the previous 12 months. The Committee is considering two alternatives for this metric: Alternative 1 Kolmogorov-Smirnov (KS) test metric The KS test metric assesses how similar the distributions of the HPL and the RTPL are over time by calculating the maximum absolute difference between the probability distributions of the HPL and the RTPL over the time series. Well modelled trading desks would be expected to feature smaller differences between the distributions. Alternative 2 Chi-squared test metric The Chi-squared test assesses similarity of distributions by dividing the time series of the HPL and the RTPL, respectively, into five bins (with each bin representing a non-overlapping range of values for the P&L measure as calculated over the time series) and counting the number of instances of the HPL and the RTPL that fall into each bin. If the HPL and the RTPL have a similar number of occurrences in each bin, they have a more similar distribution and are indicative of a well modelled trading desk. 2.1.3 PLA test failure consequences The January 2016 standard specified that trading desks that fail the PLA test would become ineligible to use the IMA and thereby be subject to capital requirements based on the standardised approach. Although this automatic consequence was intended by the Committee, the Committee acknowledges that an immediate fallback to the standardised approach from the IMA can contribute to significant volatility in the capital requirements for a given trading desk. To address concerns over volatility in capital requirements, the Committee proposes a modified, traffic light approach to smooth a trading desk s transition to the standardised approach. The proposed traffic light approach features three zones by which trading desks are categorised based on their PLA test performance. Trading desks in the green zone are those that pass the PLA test, Revisions to the minimum capital requirements for market risk 7

whereas those in the red zone are those that have failed the PLA test and must fall back to use of the standardised approach. Trading desks in the amber zone are those that have not met the full requirements of the PLA test, but that have not performed so poorly as to necessitate immediate fallback to the standardised approach. Under the Committee s proposal, the capital requirements for a trading desk in the amber zone would be subject to an additional simple, formula-based capital requirement to be added to the trading desk s IMA-based capital requirements. This additional requirement is determined by the difference between IMA and standardised approach capital requirements as determined on an aggregated level for all trading desks in the green zone and the amber zone. This difference is adjusted by (i) a weighting factor that is determined by the materiality of the amber zone trading desks in relation to the combined set of green zone and amber zone trading desks; and (ii) a multiplier that sets the severity of the capital surcharge determined as a fixed multiplier of 50%. The Committee is of the view that this additional capital requirement for trading desks in the amber zone will provide a prudent consequence for trading desks that do not meet the full requirements of the PLA test, but that perform sufficiently well so as to not necessitate exclusive use of the standardised approach. The proposed thresholds below which a trading desk would fall into the amber zone or the red zone are set out in the Annex B. Upon finalisation of the traffic light approach into the market risk standard, the Committee will continue to monitor the effectiveness of the finalised calibration of the thresholds to ensure their appropriateness. 2.1.4 Trading desk requirements As described above, the PLA test is applied to each individual trading desk. In order to promote consistency in how banks define their trading desks for this purpose, the January 2016 standard set out a number of requirements for the organisation of trading desks to which a bank intends to apply the IMA. These requirements were intended to allow flexibility so as to avoid conflict with the way banks typically organise their trading activities. The Committee has noted that some elements of the January 2016 standard s requirements for trading desks the requirement for a single head trader per desk and the restriction that a trader may only be assigned to a single trading desk could conflict with the way banks organise their trading desks. To allow more flexibility in the way trading desks are established, the Committee proposes revisions to amend these requirements as set out in Annex C. 2.2 Non-modellable risk factors The IMA model requirements of the January 2016 standard permit a bank to include a risk factor in an internal model if there are at least 24 real price observations of the value of the risk factor over the previous 12 months, with no more than a one-month gap between any two observations. The Committee intended this requirement, referred to as the risk factor eligibility test (RFET), to provide assurance that the risk factors that a bank models are sufficiently liquid and observable to be amenable to modelling. In the event that a given risk factor does not satisfy the RFET, it is classified as a non-modellable risk factor (NMRF), is to be excluded from the bank s expected shortfall (ES) model, and is subject to capital requirements determined by means of a stress scenario. Since the publication of the January 2016 standard, the Committee has received feedback from market participants that the standard is not sufficiently clear regarding (i) the meaning of real price observations and (ii) the requirements for banks use of data to calibrate internal models (eg whether the same data observations used for the RFET must be used for the calibration of banks internal models). The Committee is also aware of nascent efforts to establish data-pooling schemes that could improve the availability of real price observations for the RFET, but that may face confidentiality-driven challenges that prohibit the sharing of actual prices to subscribers of such a service. To address these issues, the Committee proposes clarifications to the RFET and a number of principles to inform assessments of the 8 Revisions to the minimum capital requirements for market risk

quality of data that banks use to calibrate their internal models. Further details are set out below, and associated revisions to the related standard text are in Annex D. A number of market participants and stakeholders have also expressed concerns that the approach defined for NMRFs may be subject to design flaws that result in disproportionately high capital requirements for some risk factors relative to the risk they pose to a bank (eg due to an arguably liquid risk factor not meeting requirements of the RFET or due to an overly conservative treatment of certain types of NMRFs). As described below, the Committee has not received compelling evidence for these issues, and seeks further feedback in response to this consultative document that could support a final decision on them. In the absence of compelling evidence, the Committee does not propose revisions to these aspects of the treatment of NMRFs. 2.2.1 Process for satisfying modellability requirements and expectations for internal model calibration The January 2016 standard required real price observations to be representative of the risk factors which are subject to the RFET. This is because risk factors for a financial instrument often do not correspond directly to observable transactions in the market. For example, a bank might model the implied volatility of an equity when modelling the risk of an equity option. Although the implied volatility of the equity is not directly observable, the bank may make a case that it can be derived from observable market transactions in other options on that equity and wish to count those as real price observations. For risk factors that can be directly observed from a market transaction, a bank may wish to model the performance of the risk factor based on available observations of a similar, but not identical, instrument and treat those as real price observations (eg to model a five-year credit default swap (CDS) spread, a bank may seek to use a CDS with a maturity close to five years as the observation for a five-year risk factor). The January 2016 standard did not elaborate the process to assess whether the observed transactions are sufficiently representative to be counted as real price observations, which has led to concerns regarding consistency in banks approaches. To facilitate consistent application of the standard, the Committee proposes to clarify the meaning of representative real price observations. It also proposes: to clarify that a bank that uses observable transactions to derive the value of underlying risk factors must establish, to the satisfaction of its national supervisor, policies and procedures to clearly set out its approach to map real price observations to the relevant risk factors; and two potential alternatives to determine how similar a risk factor of an observable transaction must be to the risk factor for a financial instrument in order to count as an observation for the RFET. The first alternative would permit a bank to establish its own ranges ( buckets ) for its risk factors within which an observable transaction may qualify as an observation for a risk factor. The ranges would be subject to certain limitations, including that a bucket can only correspond to one risk factor, and supervisory approval. The second alternative would specify buckets that banks must use. The Committee seeks feedback on the relative merits of each alternative, particularly with regard to ensuring consistency of outcomes while offering sufficient flexibility to appropriately cover the variety of instruments traded by internationally active banks. Commenters that prefer the second alternative should provide proposals for the minimum requirements for setting the buckets which are at least as granular as the buckets used in the standardised approach an example of such a structure is included in Annex D. In addition, the Committee proposes to clarify conditions whereby (i) committed quotes may be used as real price observations and (ii) data-pooling schemes could be employed to help banks satisfy the RFET. For risk factors that meet the RFET, the Committee intends to provide banks reasonable flexibility in the choice of data to be used to calibrate internal models. However, it is the responsibility of the bank to use appropriate data to ensure its model provides a representative measurement of risk. To ensure that Revisions to the minimum capital requirements for market risk 9

the data used for model calibration are consistently robust across banks, the Committee proposes a number of principles to which banks will be expected to adhere when selecting data to calibrate their models. 2.2.2 Impact of the NMRF framework on seasonal markets The Committee has received feedback from market participants that the RFET requirement for there to be a gap of no longer than one month between any two price observations may result in some risk factors being inappropriately deemed as non-modellable despite other evidence of market liquidity. In particular, challenges in identifying observable prices during certain periods of the year have been cited due to seasonality of markets and low volumes of trading during holiday periods. Although the Committee does not propose any changes to the requirements of the January 2016 standard in this regard at this time, the Committee welcomes data and comments to confirm the validity and materiality of these concerns. Respondents should provide concrete evidence and data based on specific examples of risk factors that are considered to have demonstrated adequate liquidity and observability in stress periods but are unable to meet the requirements of the RFET due to the above concerns. The Committee also welcomes alternative proposals for the RFET requirement based on the identified examples. In the absence of compelling evidence, the Committee will not make changes to the treatment of NMRFs in this regard. 2.2.3 Impact of NMRF idiosyncratic equity risk Per the January 2016 standard, capital requirements for NMRFs are calculated using a stress scenario for each type of NMRF, with the resulting stress losses to be aggregated without recognition of diversification benefits. As an exception, for NMRFs associated with idiosyncratic credit spread risk, banks are permitted to use a single stress scenario and may recognise diversification. The Committee specified this exception for credit risk due to a view that failure to do so would result in disproportionately high capital requirements for these idiosyncratic credit risk factors. The Committee has received feedback that, for some banks, the lack of a corresponding exceptional treatment for idiosyncratic equity risk is also resulting in exceedingly high capital requirements for equity risk NMRFs. Due to insufficient evidence on the potential materiality of this issue, the Committee does not propose to change the treatment of idiosyncratic equity risk at this time. However, the Committee welcomes comments on the materiality of this issue (and associated supporting concrete evidence and data) to support its consideration. The Committee also welcomes feedback on whether a revision per the amended text of paragraph 190 as included in Box 3 below could address the issue. In the absence of compelling evidence, the Committee will not make changes to the treatment of NMRFs in this regard. 10 Revisions to the minimum capital requirements for market risk

Box 3 190. Each non-modellable risk factor is to be capitalised using a stress scenario that is calibrated to be at least as prudent as the expected shortfall calibration used for modelled risks (ie a loss calibrated to a 97.5% confidence threshold over a period of extreme stress for the given risk factor). For each non-modellable risk factor, the liquidity horizon of the stress scenario must be the greater of the longest time interval between two consecutive price observations over the prior year and the liquidity horizon assigned to the risk factor in paragraph 181. For non-modellable risk factors arising from idiosyncratic credit spread risk or idiosyncratic equity risk arising from spot, futures and forward prices, equity repo rates, dividends and volatilities, banks may apply the same stress scenario. Additionally, a zero correlation assumption may be made when aggregating gains and losses provided the bank conducts analysis to demonstrate to its supervisor that this is appropriate for example, analysis on the residuals and showing that residual distributions from factor models are homogeneous within each residual distribution and heterogeneous against others, and do not exhibit serial correlation in the time series. Footnote No correlation or diversification effect between other non-modellable risk factors is permitted. In the event that a bank cannot provide a stress scenario which is acceptable for the supervisor, the bank will have to use the maximum possible loss as the stress scenario. The aggregate regulatory capital measure for I non-modellable idiosyncratic credit spread risk factors that have been demonstrated to be appropriate to aggregate with zero correlation, J nonmodellable idiosyncratic equity spread risk factors that have been demonstrated to be appropriate to aggregate with zero correlation and K risk factors in model-eligible desks that are non-modellable (SES) is: 2 SSSSSS = IIIIIIII NNNN,ii II ii=1 JJ 2 + IIIIIIII NNNN,jj jj=1 KK + SSSSSS NNNN,kk kk=1 where ISES NM,i is the stress scenario capital charge for idiosyncratic credit spread non-modellable risk i from the I risk factors aggregated with zero correlation; ISES NM,j is the stress scenario capital charge for idiosyncratic equity non-modellable risk j from the J risk factors aggregated with zero correlation; and SES NM,k is the stress scenario capital charge for non-modellable risk k from K risk factors. Footnote The tests are generally done on the residuals of panel regressions where the dependent variable is the change in issuer spread while the independent variables can be either a change in a market factor or a dummy variable for sector and/or region. The assumption is that the data on the names used to estimate the model suitably proxy the names in the portfolio and the idiosyncratic residual component captures the multi-factor name basis. If the model is missing systematic explanatory factors or the data suffer from measurement error, then the residuals would exhibit heteroscedasticity (which can be tested via White, Breuche Pagan tests, etc) and/or serial correlation (which can be tested with Durbin Watson, LM tests, etc) and/or cross-sectional correlation (clustering). 3. Scope of market risk capital requirements The January 2016 standard set out a revised definition of the scope of the market risk capital requirements. This included a new definition of the boundary between a bank s trading book and its banking book, with the former to be subject to capital requirements under the market risk standard, and revisions to the standards for structural FX positions (ie FX positions that hedge a bank s capital ratio and may therefore be exempted from FX capital requirements). Revisions to the minimum capital requirements for market risk 11

The January 2016 standard s boundary definition was intended to address shortcomings identified with the Basel II boundary definition by reducing the ability of banks to arbitrage the different capital requirements between the trading book and the banking book, and facilitating consistent implementation of the boundary across banks. The January 2016 standard s treatment of structural FX positions was intended to remove redundant text used in the Basel II market risk framework and therefore simplify this part of the framework, but not to materially alter the current practices applied by banks under Basel II framework. Based on its monitoring since the publication of the standard, the Committee has identified areas of both of the above elements that it believes require clarification in order to ensure they continue to have their intended impact and can be implemented consistently across banks. Further details on the proposed revisions are described below, and associated revisions to the related accord text are in Annex E. 3.1 Treatment of structural FX positions Instruments with FX risk are subject to market risk capital requirements regardless of whether they are held in the trading book or the banking book. However, supervisors may permit banks to exclude certain FX risk positions from the calculation of net open currency positions if the position is entered into to completely or partially hedge against adverse effects on the bank s capital ratio due to changes in exchange rates. In the January 2016 standard, this exclusion was limited to the maximum of the amount of investments in consolidated subsidiaries or non-consolidated affiliates. The Committee proposes revisions to allow the amount of structural FX positions that may be exempted from market risk capital requirements to be measured based on the FX risk stemming from an investment, rather than the amount of investment itself. The limit on the amount of such exempted positions would be the amount of the risk position that neutralises the sensitivity of the bank s capital ratio to movements in exchange rates. In addition, the Committee proposes revisions to clarify that structural FX positions in foreign branches of a bank can be included in the scope of the structural FX exemption. The Committee believes these proposed revisions would enhance consistency of the treatment of structural FX positions across jurisdictions while more closely aligning the proposed standard with the treatment currently applied, as was originally intended in the January 2016 standard. 3.2 Boundary between the trading book and the banking book The January 2016 standard definition of the boundary between the trading book and the banking book specified types of financial instruments that must be in the banking book and similar instruments that must be in the trading book. It also specified financial instruments that are expected to be in a particular book but could be designated to a different book with supervisory approval. This additional detail, not present in the Basel II market risk framework boundary definition, was intended to promote consistent implementation of the boundary, and limit the ability of banks to arbitrage capital requirements by choosing, without restriction, to which book they designate instruments. The Committee has identified that in some cases financial instruments can be both in the list of instruments that must be in a particular book, and in the list that are expected to be in the other book. In these cases, it may not be clear which requirement takes precedence. The Committee therefore proposes amendments to this part of the standard to clarify the approach in these situations. The proposed revisions also clarify under what conditions equity investments in funds (eg exchange-traded funds) can be included in the trading book. Under the proposals, banks may assign to the trading book funds: (i) for which daily price quotes are available; (ii) which track a non-leveraged benchmark; and (iii) which demonstrate a tracking difference, ignoring fees and commissions, for which the absolute value is less than 1%. 12 Revisions to the minimum capital requirements for market risk

4. Simplified alternative to the standardised approach In June 2017, the Committee published a consultative document to propose a simplified alternative to the standardised approach to market risk capital requirements. 8 The proposal to include such a simplified alternative was intended to facilitate adoption of the market risk standard for banks other than those that are internationally active. The consultative document proposed that the simplified alternative could take the form of one of the following: (i) a reduced form of the January 2016 standard s sensitivities-based method; or (ii) a recalibrated version of the Basel II standardised approach. In response to comments received, the Committee is of the view that a recalibrated Basel II standardised approach would be better suited to facilitate the adoption of the standard by the banks for which a simplified alternative is intended. Annex F sets out the proposed recalibration of the Basel II standardised approach. To recalibrate the approach, the Committee proposes to apply a multiplier to the capital requirements in each risk class of the Basel II standardised approach. No other amendment to the approach is proposed. The multipliers proposed are set out in Table 1. The final calibration of these multipliers will be determined based on further analysis of impact data provided by banks, and feedback provided to this consultative document. Proposed risk class multipliers for the simplified SA Multiplier to be applied to Basel II standardised approach General and specific interest rate risk 1.50-2.00 General and specific equity risk 3.00-3.50 Commodity 1.50-2.50 FX 1.25-1.50 Table 1 The recalibration proposed is intended to make the Basel II standardised approach s calibration comparable with, but slightly more conservative than, the revised full standardised approach. Given its relatively more conservative proposed calibration, the Committee does not propose to specify eligibility requirements for banks that may use this approach. Nevertheless, the Committee notes that the simplicity of the approach means that it may not be appropriate for banks that (i) are globally systemically important banks (G-SIBs); (ii) use internal models for determining the market risk capital requirements for part of their trading book; or (iii) maintain correlation trading portfolios. Next steps The Committee welcomes comments on all aspects of these proposals. The Committee s scope of material potential revisions to the revised market risk framework are limited to those included in this consultative document. Respondents are requested to limit their feedback to views on the proposals contained herein. Comments should be uploaded at www.bis.org/commentupload.htm by 20 June 2018. All comments will be published on the website of the Bank for International Settlements unless a respondent requests confidential treatment. 8 Basel Committee on Banking Supervision, Simplified alternative to the standardised approach to market risk capital requirements Consultative Document, June 2017, www.bis.org/bcbs/publ/d408.pdf. Revisions to the minimum capital requirements for market risk 13